Spectral Instruments Series 800, User Manual, Page: 38 / 70

Spectral Instruments
Pt # 2500- 38
4.3.2 How
The three calibration constituents must each be manipulated differently. The
master bias is used as is. The master dark image must be multiplied by the ratio
of the exposure time of the target image to the exposure time of the master dark
image. For most imaging systems this scaling is by integer values, as floating
point images are not usually used. SI Image uses floating point so fractions are
not a problem and neither are negative numbers.
When you actually apply a flat field image correction, the target image must first
be multiplied by some constant before it is divided by the flat image in order to
preserve the significance in integer format images. A good choice is the average
value of a typical "bright" region in the master flat image. Select such an area in
the master flat, set a region of interest and determine the average value in this
region of interest. Record the value of this average from the ROI in master flat
image “bright area”. This average is the multiplier or “scaling parameter” you
will use to scale all target images that will be corrected using this master flat
image.
You have a master bias, a master dark of some exposure time that has been
corrected for hot pixels and a master flat from which you got a scaling parameter.
You are ready to proceed to correct a target image.
Bring up the target image and subtract the master bias. Multiply the master dark
by the ratio of the exposure times and subtract it from the target image. Multiply
the target image that has now been bias and dark corrected by the scaling
parameter obtained above. This step requires that the result be an extended
precision image. Either floating point or signed long will suffice.
Now divide the scaled target image by the master flat image. This step requires
promoting the master flat image to the same type as you selected for the scaled
target scene image before the division. It may also require a demotion of the
result to a shorter word-length afterwards. Again, in SI Image, these tasks are
automated for you.
The result is a new version of the target image where the shading pattern is
removed. This is a flat-fielded target image.
4.3.3 Limitations On The Flat Field Process
Effective flat field correction depends upon stability of the illumination and the
attenuation. CCDs are strongly wavelength sensitive. The quantum efficiency
variation across the sensor is different for different wavelengths of incoming
light. There for, flat field images vary with the color of the incoming light –
especially if that light is in a very limited wavelength band. The target images
and the master flat images must be exposed to nearly the same color of light.